Artificial Intelligence used to decode animal language

Research programs are working on the design of an “interspecies phonetic alphabet”, developed with artificial intelligence models.

Researchers have been working on machine learning models for around fifteen years (machine learning), to break down the practice of a language into a juxtaposition of words. And obtain translations or sentence constructions by arranging these words.

In particular by taking into account the context, to avoid misunderstandings or false friends. And with rules for arranging words, born from the grammar specific to each idiom. This is how we developed automatic translation software, which is becoming more precise.

Several scientific programs are dedicated to animal language. Like for example the “Earth species” project. Three of its experts published this year a research document at Cornell University, in the United States, devoted to an idea that would be an “interspecies phonetic alphabet”, which would be used to transcribe in the form of texts, the sounds generated by animals.

Scientists want to establish a transcription system that is:

• accurate – which retains as much information from the original audio recording as possible,
• concise – to express information with a minimum of tokens (tokens) to have AI models that do not consume too much technical resources.
• interpretable, that is to say that it is at some point understandable by humans.

We must therefore deal with these three criteria to develop a methodology that works. And that’s the whole difficulty. Especially when we consider the great variety of the animal world.

This is the reason why, in order to restrict their field of action, researchers specialize in certain animals. Like the CETI (Cetacean Translation Initiative) program founded as an NGO in 2020, based on work carried out since the 1960s. It is dedicated exclusively to cetaceans. The team brings together specialists in marine biology, cryptographers, roboticists and acousticians.

Each of these research teams is fully aware that the available technology accelerates a possible outcome. And this now raises the question of the consequences of such an innovation. Particularly in terms of impact on animal welfare.

Mastery of interspecies communication, placed in the wrong hands, can have dramatic effects. For example, to facilitate the capture or exploitation of wild species. With, as a result, disturbances to the environment in general, the scale and effects of which we have not yet measured.

It is therefore not only a subject for technicians, but it requires a collective ethical approach to the impact of such a discovery. This reflection must accompany this research work on interspecies dialogue.